ESSAY // THE READING ROOM

The encryption checkbox

written 2026-11-02 · about 2 min read · 6 paragraphs

A control on the contact form claimed to encrypt the message. It encoded it in base64, named a library that was never installed, and sent plaintext anyway.

The closing chapter of this site has a one-line dispatch console: you type an address and a message, it posts to /api/contact, and the message lands in an inbox backed by Postgres. For a while it also had a checkbox that said the message would be encrypted.

What the checkbox actually did was wrap the text in a header that imitated the shape of an encrypted payload and paste a btoa() call inside it. That is three separate claims failing at once. Base64 is an encoding, not an encryption — anyone who sees the string can undo it. The key library named in the label was never added to package.json, so there was no key to speak of. And the endpoint received the body as plaintext regardless, which means the last hop of the story the label told was simply false.

It was removed rather than reworded. Commit 8d5819b dropped the fake PGP claim, and the comment that replaced it in ChapterPaddock.tsx says the claim must not come back before there are keys and real encryption on both ends of the wire.

The general rule is narrower than it looks: a control is a statement about the system. A checkbox, a badge, a lock icon, a rate-limit note — each one asserts something a reader will not test before trusting. The bar is not that the claim be defensible in a legal sense; it is that every hop the claim describes is actually doing what the label says.

That is also why the rest of this site measures itself in public. The telemetry page names its window, its polling interval and its rate limit; the repository index names the API call its numbers came from; the verifier suite fails a build when a printed figure has no source response behind it. Not because visitors ask for it, but because the alternative is a page full of controls that describe a system that does not exist.

Real encryption at both ends is still not here. It is a decision with a cost — a key management story, an inbox that has to decrypt, a rate limit worth defending — and this page would rather leave the box uninstalled than install it again as decoration.

WHAT THIS TEXT POINTS AT

Paths are relative to the repository root; a bare hash is a commit in its history. Every claim above can be opened and contradicted.

  • nextjs-frontend/src/components/Editorial/ChapterPaddock.tsx
  • nextjs-frontend/next.config.ts
  • go-backend/main.go
  • 8d5819b